An Experimental Study Of Human And Algorithm Trader Performance

Trading algorithms are an increasing presence in many markets. Some researchers have focused on the formulating trading algorithms with the benchmark of outperform-ing human traders. A study by Das et al.[1] undertook this challenge in one of the most robustly competitive scenarios possible——experimental Continuous Double Auctions(CDA). They showed that as a simple algorithm that make assessments and action faster their relative performance increases and eventually exceeds that of humans with the same valuations. We showed their design with equal numbers of algorithms and humans on each side of the market is a very special case. We also show that under competitive market, monopoly market and duopoly market when humans on one side of the market and the algorithms on the other, increasing algorithm speed leads increasing poor performance. When we consider monopoly and duopoly market structures with humans and algorithms on opposing market sides, increasing algorithm speed leads to market outcomes increasingly close to perfect first degree price discrimination with algorithms accruing zero gains from exchange.